#!/usr/bin/env python3 # Copyright 2020 NLP Odyssey Authors. All rights reserved. # Use of this source code is governed by a BSD-style # license that can be found in the LICENSE file. import torch FLOAT_DTYPES = [ torch.float16, # or torch.half torch.float32, # or torch.float torch.float64, # or torch.double torch.bfloat16 ] INT_DTYPES = [ torch.int8, torch.int16, # or torch.short torch.int32, # or torch.int torch.int64, # or torch.long ] def main(): for proto in range(1, 6): for use_zip in [False, True]: for dtype in FLOAT_DTYPES: save([1.2, -3.4, 5.6, -7.8], dtype, proto, use_zip) for dtype in INT_DTYPES: save([1, -2, 3, -4], dtype, proto, use_zip) save([1, 10, 100, 255], torch.uint8, proto, use_zip) save([True, False, True, False], torch.bool, proto, use_zip) def save(data, dtype, proto, use_zip): str_dtype = str(dtype)[6:] str_zip = '_zip' if use_zip else '' torch.save( torch.tensor(data, dtype=dtype), f'tensor_{str_dtype}_proto{proto}{str_zip}.pt', pickle_protocol=proto, _use_new_zipfile_serialization=use_zip) if __name__ == '__main__': main()